Using molecular and morphometric data as operational criteria for the analysis of a threatened rainforest species complex shows interspecific variation, with implications for cryptic-species delimitation and conservation

Study taxa

Current taxonomic classifications of Fontainea are based on traditional morphologically focussed techniques. Several taxa display unique identifying features that allow for simple delineation between the species following this approach (Jessup and Guymer 1985; Forster 1997; Forster and Van Welzen 1999). However, morphological plasticity within Fontainea from the southern clade (Rossetto et al. 2000) suggests that a clearer resolution of the genus could be achieved by combining analysis of molecular markers and morphological characters to evaluate phenotypic variation among putative taxa. Recent evidence from a set of reduced-representation markers (DArTseq) showed F. oraria to be closely aligned with several F. australis populations, and that F. oraria potentially represents an ecotype that forms part of a broader species continuum within F. australis (Brunton et al. 2024). To provide a more definitive characterisation of the species boundaries, a rigorous analysis with phenotype data was recommended for future studies of this challenging complex.

Sampling

Our combined analysis of silica-dried leaf material to generate reduced-representation genomic data for molecular analysis, along with a separate set of oven-dried material from field-collected specimens for the morphometric component. Leaf material from populations of F. australis, F. oraria and the Coffs Harbour Fontainea was collected throughout their geographical distributions in central-eastern Australia (Fig. 1).

Fig. 1.

Sampling locations of an Australian rainforest complex, including Fontainea australis (see middle inset map for population detail), F. oraria and F. sp. Coffs Harbour.

In total, 186 genetic accessions and 152 leaf accessions from 14 populations of the 3 Fontainea taxa were collected for genetic and morphological analyses (Table 1). In addition, representatives from two F. rostrata populations from the Gympie region in Queensland (Qld; Goomboorian and Greens Creek) were used as an outgroup to provide a measure of support for species delineation for genetic analyses. Fontainea rostrata was selected because it was previously recovered as a distinct species in a sister subclade to F. australis and F. oraria. (Brunton et al. 2022). A voucher representative for each population was lodged with the Queensland Herbarium (BRI), or where samples were supplied by an external source (F. australis – CC, CRY), we refer to the NSW Herbarium (NSW) voucher details (Table 1).

SNP genotyping and quality filtering

Genotype data were generated from silica-dried leaf samples (10–15 mg per individual) sent to Diversity Arrays Technology Pty Ltd (DArT, Canberra, ACT, Australia) who performed all DNA extractions and raw sequencing by using proprietary analytical pipelines, specifically employing DArTseq analysis. DArTseq utilises a combination of complexity reduction through restriction enzymes, implicit fragmentation size selection, and next-generation sequencing (NGS) (Kilian et al. 2012), similar to double-digest restriction-associated DNA sequencing (ddRAD; Peterson et al. 2012). As a result, DArTseq produces a collection of co-dominant SNPs accompanied by relevant metadata, providing insights into the quality of each SNP. The metadata include a reproducibility score for each locus, indicating the consistency of genotype calls across replicates.

DArTseq has been successfully applied to define a range of taxonomically challenging plant groups (Sansaloni et al. 2011; Steane et al. 2011; O’Donnell et al. 2023), including Fontainea (Brunton et al. 2022). Also, notably, the chromosome count and ploidy status of Fontainea are currently unknown and have not been investigated in the study species.

For this study, we used a dataset similar to that of Brunton et al. (2024); however, this new SNP data included samples of the novel Fontainea entity from Coffs Harbour. The quality of the SNP dataset was assessed using filtering scripts from the custom R package ‘RRtools’ from the Royal Botanical Garden, Sydney (Rossetto et al. 2019). To ensure the reliability of the analyses, only high-quality markers meeting specific criteria were retained. SNPs with a reproducibility of ≥96% and less than 20% missing data were included. In cases where a marker had multiple SNPs, one SNP was randomly chosen to avoid linkage bias. The proportion of missing data for the samples was 3.74% after quality checking. After these steps, the dataset was reduced to 18,314 SNPs that were used for subsequent genetic analyses.

Genomic-data analysis

Genetic trees constructed for F. australis, F. oraria and the Coffs Harbour Fontainea were rooted using Fontainea rostrata because this species is related to, but distinct from, the three focal groups within this southern complex (Brunton et al. 2022). Although this phylogeny-reconstruction approach considers the numerous independently inherited SNPs as sharing a single gene tree, which may not fully account for their independent evolution and inheritance, it still holds potential to offer insights into the phylogenetic relationships among individuals within the Fontainea complex, provided the interpretation is approached with caution.

To construct the genetic trees, a NEXUS file was generated and exported from RStudio (ver. 4.3.3, Posit Software, PBC, Boston, MA, USA, see https://posit.co/products/open‐source/rstudio/) by using a down-sample approach (SVD quartets) to randomly sample the filtered SNP dataset so that all classes have the same frequency as the minority class (Chifman and Kubatko 2014). Phylogenetic relationships among the three species were then reconstructed using FASTA alignments generated using Decipher (Wright 2016). To identify the best-fit model to implement for the phylogenetic analysis, we used the modelTest function from the R package Phangorn (ver. 2.11.1, see https://cran.r-project.org/package=phangorn; Schliep 2011) to highlight the substitution model with the greatest statistical support (GTR + G) on the basis of the lowest values of Akaike information criterion (AIC), AIC adjusted for small sample sizes (AICc) and Bayesian information criterion (BIC). Maximum-likelihood (ML) topologies were then constructed using Phangorn (ver. 2.11.1; Schliep 2011) and Phytools (ver. 2.1-1, see https://cran.r-project.org/package=phytools/; Revell 2012). To infer confidence values on the ML trees, we applied 100 replicates for the DArTseq dataset, then plotted bootstrap support values on tree edges for the GTR + G. Last, an optimised, GTR + G tree was run to allow for stochastic rearrangements. Optimised trees were then exported in Newick format for additional editing and visualised using FigTree (ver. 1.4.4, A. Rambaut, see http://tree.bio.ed.ac.uk/software/figtree/).

SplitsTree (ver. 4.41.6, see https://github.com/husonlab/splitstree4; Huson 1998) was used to conduct a network analysis with a NeighbourNet tree by using the same NEXUS file that was used for the phylogenetic analysis. Default settings on uncorrected characters with ambiguous states were applied to show a visual summary of genetic relationships and to identify areas that could reflect reticulation, incomplete lineage sorting or other processes that are represented through dense networks or ‘webs’ among branches.

To explore patterns of genetic structure among Fontainea taxa, we conducted a principal-component analysis (PCA) by using adegenet (ver. 2.1.10, see https://cran.r-project.org/package=adegenet; Jombart 2008) and generated a pairwise population genetic differentiation (F_ST) matrix with the diveRsity package (ver. 1.9.90, see https://cran.r-project.org/package=diveRsity; Keenan et al. 2013). To determine the correlation between genomic distribution and geographic distance (isolation by distance, IBD), we performed a Mantel test in the Vegan package (ver. 2.6-4, J. Oksanen et al., see https://cran.r-project.org/package=vegan/) by using pairwise Euclidian-distance matrices.

Finally, we assessed genomic diversity within the Fontainea complex by using the Hierfstat package (ver. 0.5-11, see https://cran.r-project.org/package=hierfstat; Goudet 2005) to estimate allelic richness (A_R) with a rarefaction method, number of alleles per locus (size), observed heterozygosity (H_O), expected heterozygosity (H_E), inbreeding coefficient (F_IS) and number of private alleles (n_pa).

Leaf morphometric characters and multivariate analysis

To examine the congruence of morphological traits with genetic divergence, we evaluated a set of seven leaf characters. Leaf-sampling protocols are not always clear in the literature; however, we followed similar published methods for morphometric (Silva et al. 2012) and functional analyses (Pérez-Harguindeguy et al. 2013), whereby medium-aged to young, fully expanded leaves were collected from the most sunlit areas of the canopy of each individual plant (>1.5 m in height). Morphological characters analysed include those reported in taxonomic treatments (Jessup and Guymer 1985; Forster 1997), as well as characters that represent key ecological strategies among plants and that appeared variable on the basis of our field observations (Fig. 2).

Fig. 2.

Leaf morphological diversity of the Fontainea complex from central eastern Australia. Each panel (a–h) displays a group of three oven-dried, medium-aged to young, fully expanded leaves collected from the canopy of a representative individual from various populations across the following three Fontainea taxa: (a) F. australis from private property (population – EHR) adjoining Springbrook National Park, Qld, (b) F. australis from private property in Mooball, NSW (population – KR), (c) F. australis from Limpinwood Nature Reserve (population – LW), NSW, (d) F. australis from road reserve in Mooball, NSW (population – MRR), (e) F. oraria from private property in Lennox Head, NSW, (f) F. oraria from private property in Lennox Head, NSW, (g) F. sp. Coffs Harbour from Korora, NSW, (h) F. sp. Coffs Harbour from wesern Korora, NSW. Scale bar: 20 mm.

All characters were measured on the abaxial surface, with the average over three leaves collected for each specimen after being pressed flat while oven-dried in paper bags at 70°C for 72 h. This average-estimate approach follows the methods of a number of similar leaf morphometric studies (Rumpunen and Bartish 2002; Viscosi et al. 2009; Viscosi and Cardini 2011). Leaf length (LL) was measured at the midrib from the leaf base to apex, and leaf width (LW) at the widest point. Petiole length (PL) was measured as the distance from the apex to the base where the leaf was removed from the plant. For a quantitative leaf-shape surrogate, we used leaf ratio (LRTO), which was calculated as LL:LW. Averages of these measurements were calculated using ImageJ (Schneider et al. 2012), which was also used to calculate leaf area (LA).

Dry leaf mass (DLM) was measured from the oven-dried weight (g); this measure has been closely linked to absolute water content and leaf form and structure, which contribute to the carbon-integration cycle for leaf expansion and leaf-mass accumulation (Chaves et al. 2002; Ullah et al. 2013). Leaf water content state is strongly associated with leaf growth, turgor, transpiration, photosynthesis and stomatal conductance (Kramer and Boyer 1995). Typically, larger leaves are related to wet, hot and intense-light environments, whereas plants with small leaves are mostly found at high elevations or latitudes or in arid, hot systems (Wright et al. 2017). Specific leaf area (SLA; leaf area:dry leaf mass) was included as a key functional trait that represents the available area for capture of solar irradiance (Milla and Reich 2007) and has also been linked to important biological traits, including leaf nitrogen concentration, plant growth rate and lifespan (Lambers and Poorter 1992; Wright et al. 2004). In addition, interspecific differences of SLA have been strongly associated with resource partitioning among species, with low-SLA species related to nutrient-poor soils and a slow-growth strategy (Poorter and De Jong 1999; Baraloto et al. 2006).

Although floral and fruit features are an important diagnostic feature for taxonomic treatments of Fontainea (Jessup and Guymer 1985; Forster 1997), the asynchronous and irregular phenology, both within and among species, limited collection to only a few populations and individuals. We therefore could not include such data in the morphometric analysis.

As such, we applied a multivariate approach to assess leaf morphological differentiation among the three Fontainea taxa using the R package MorphoTools2 (ver. 1.0.1.1, see https://cran.r-project.org/package=MorphoTools2; Šlenker et al. 2022). Morphometric data were evaluated, first, by using a PCA based on the seven leaf traits, because this identifies potential group structure without assigning a priori groups for the samples (Šlenker et al. 2022). Furthermore, a canonical discriminant analysis (CDA) and classificatory discriminant analysis were performed at multiple levels (population and taxon) by using a linear discriminant analysis (LDA), with a stepwise selection to assess the marginal effect of certain characters with a leave-one-out cross-validation. We also performed the non-parametric, k-nearest neighbours’ (KNN) classification analysis to assess classifications at the individual level.

Last, to determine whether discrete leaf characters could discriminate among taxa, we generated conditional-inference trees with the R package partykit (ver. 1.2-20, see https://cran.r-project.org/package=partykit; Hothorn and Zeileis 2015) by using a classification tree analysis (CART). As an additional comparison to which leaf traits could be used to classify the three species, we also fit ANOVA models for the three taxa with the base R function, aov. Individual differences among taxa were determined by running Tukey HSD post hoc tests on the ANOVA model fit, with multiple comparisons of means at a 95% family-wise confidence level.

Phylogenetic associations, genetic structure and diversity

The concatenation of the SNPs detected using DArTseq resulted in an alignment length of 18,457 nucleotides by using the quality-filtered data that also included F. rostrata. Analysis of the best-fit substitution model for phylogenetic reconstruction by using the DArTseq SNPs identified the GTR + G model as the best-fit model (Supplementary Table S1), which produced an ML tree with strong support for the delineation of the major clades. The model-based ML tree recovered a topology to support F. australis populations as divergent from F. oraria and the Coffs Harbour Fontainea entity (Fig. 3a). For F. australis, the ML tree was congruent with previous phylogenetic associations (Brunton et al. 2024); however, our revised analysis showed that three geographically proximate populations of MNP, KR and MRR as well as WG and HB were strongly supported as a well-differentiated genetic subgroup that is distinct from the remaining nine F. australis populations in a branch that was sister to F. oraria (Fig. 3a). There was also strong support (bootstrap value = 100), indicating that the Coffs Harbour entity is a distinct genetic cluster. These genetic associations were also reflected in the consensus network tree (Fig. 3b), which supported the novel Fontainea from Coffs Harbour as a genetically distinct group.

Fig. 3.

(a) Phylogenetic tree and (b) a consensus network of a Fontainea complex from central-eastern Australia constructed from individually aligned and concatenated DArTseq SNP sequence markers (18,457) by using a maximum-likelihood analysis. Fontainea rostrata was used as an outgroup representative to root the tree. The values on the phylogenetic tree at the branch nodes represent bootstrap support on the basis of the best-fit GTR + G model. Fontainea lineages recovered are represented outside the tip labels of the tree and tips and branches are coloured by species.

Network analysis concentrated on the three focal Fontainea groups recovered relationships similar to the phylogenetic topology and broader network with F. rostrata, with three main clusters (Fig. 4a), namely, one with Coffs Harbour Fontainea, a second with F. oraria, and a third, dense clade of all the F. australis samples comprising a range of subclusters that mostly represent the geographically clustered populations. PCA ordination (Supplementary Fig. S1) demonstrated similar patterns of genetic clustering that also identified Coffs Harbour Fontainea as a distinct group and F. oraria as closely aligned with some F. australis populations (MRR, MNP, KR, WG). IBD analysis showed that genetic divergence was strongly associated with geographic distance (Fig. 4b). This pattern was also reflected in the pairwise population estimates of genetic differentiation (F_ST) for the three Fontainea groups (Table S2). F_ST values among F. australis ranged from a level of minimal differentiation of F_ST = 0.232 (CRY) to a maximum of F_ST = 0.352 (MNP), with an overall average F_ST value of 0.292 (Table S2). Fontainea oraria had an average pairwise population F_ST value of 0.376 compared with the F. australis populations. Coffs Harbour Fontainea had an average pairwise population F_ST value of 0.677 compared with F. australis, and 0.685 with F. oraria (Table S2).

Fig. 4.

(a) Relationship between genetic and geographic distribution (isolation by distance, IBD) for three Fontainea taxa, F. australis, F. oraria and Coffs Harbour Fontainea (see inset legend), by using a Mantel test and, (b) NeighbourNet analysis of the three Fontainea taxa (see inset legend) from the rainforests of central-eastern Australia, generated using SplitsTree.

Genomic-diversity data showed similar results across most parameters for the F. australis populations; however, the inbreeding coefficient (F_IS) was elevated among some populations, such as CRY and WG (Table 2), indicating high kinship levels, even though it has a relatively high number of private alleles compared to the other Fontainea populations. Despite its small population size, and isolated locality, F. oraria showed relatively high diversity values (A_R, H_O, H_E) and a negligible inbreeding coefficient (Table 2). Coffs Harbour Fontainea showed considerably lower values of genomic variation (A_R, H_O, H_E) than did F. australis and F. oraria, with a relatively large number of private alleles (n_pa), especially when compared with most F. australis populations (Table 2).

Morphometric analysis

PCA analysis of leaf traits without predefined group assignments did not show any clear group structure among the three Fontainea groups on the basis of the seven leaf traits (Fig. 5a). By contrast, the CDA of leaf characters showed distinct clustering of this Fontainea complex (Fig. 5b). There was clear separation of F. australis and the putative new entity, Coffs Harbour Fontainea, along the first axis, and the CDA also illustrated some overlap in morphometric characters between F. australis and F. oraria (Fig. 5b).

Fig. 5.

Principal-component analysis (PCA) of two Fontainea species and the undescribed entity from Coffs Harbour performed using seven leaf traits and without prior group assignment (a) and a Canonical discriminant analysis (CDA) of three Fontainea taxa (aust, F. australis – red; coffs, Coffs Harbour Fontainea – green; or, F. oraria – blue) highlighting morphometric cluster groupings on the basis of seven leaf traits with a leave-one-out cross-validation (b).

Generally, the series of classification analyses showed similar results and a good ability to classify the three groups within this Fontainea complex at multiple levels by using an LDA and KNN approach. At the population level (Table 3), several F. australis samples were classified with F. oraria, including some individuals from the Mooball region (KR, MRR) and the WG population. LDA at the taxon level (Table S3) classified ~9% of F. australis as F. oraria, and one sample each of Coffs Harbour Fontainea and F. oraria as F. australis.

LDA results at the taxon level also showed a high proportion (>90%) of accurate classifications for F. australis and F. sp. Coffs Harbour and classified one of the seven F. oraria samples with F. australis (Table S3). KNN analysis showed that the highest number of correct classifications was at k = 4 (Fig. S2) and reflected the LDA results, with some F. australis individuals classified as F. oraria, several F. sp. Coffs Harbour individuals classified with F. australis, and >85% of F. oraria classified as F. australis (Table 4).

Conditional-inference trees showed the three Fontainea groups were best characterised by two (SLA, DLM) of the seven leaf traits used in the analysis. Fontainea sp. Coffs Harbour was best characterised by low SLA, and at low SLA, is distinguished from some F. australis classifications by high DLM (>0.351, Fig. 6). For the larger SLA (>0.027–0.065), all samples were classified as F. australis, but with a 3.4% error, with a small number of samples each from F. oraria and Coffs Harbour Fontainea being misclassified (Fig. 6). For SLA >0.065, there was a roughly equal classification between F. australis and F. oraria, albeit with a small number of samples (n = 9).

Fig. 6.

Conditional-inference tree examining three Fontainea taxa (aust, F. australis; coffs, F. sp. Coffs Harbour; or, F. oraria), showing specific leaf area (SLA) and dry leaf mass (DLM) as the best traits to distinguish among species from seven leaf characters used as explanatory variables for classification.

In addition to the conditional inference-tree results of SLA and DLM as key traits to classify among the three Fontainea groups, significance tests showed that three morphological characters (PL, SLA and DLM, Fig. 7a–c) were able to distinguish among the groups (P ≤ 0.05), namely, PL (P = 0.0417), SLA (P ≤ 0.0001) and DLM (P ≤ 0.0007). Variation among groups was supported in post hoc calculations of the mean 95% confidence interval for each significant trait (Table S4). These showed that for SLA, there was significant variation among all three Fontainea taxa (P ≤ 0.001). For DLM there was also significant (P = 0.008) difference among Coffs Harbour Fontainea, F. oraria and F. australis. However, there was a significant difference for PL between only Coffs Harbour Fontainea and F. australis. The remaining leaf characters (LL, LW, LA, and LR) showed no significant variation among species, although a weak relationship (P = 0.098) for LRTO was detected (Fig. S3).

Fig. 7.

(a–c) Box plots representing three significantly different leaf traits, namely, petiole length (PL), specific leaf area (SLA) and dry leaf mass (DLM), among three, threatened Fontainea species (aust, F. australis; coffs, F. sp. Coffs Harbour; or, F. oraria) from central-eastern Australia.

Genetic support of described species boundaries

Our extensive exploration of this Fontainea complex identified several distinct clusters, and delimitation among them on the basis of genomic characters. In relation to the genetic–species boundaries of F. australis and F. oraria, our analyses showed that the phylogenetic and morphological distinction between some F. australis populations and F. oraria are not so well defined. Our SNP-based phylogeny showed that F. australis and F. oraria share close affinities. The phylogeny highlighted that the most easterly F. australis populations from the Mooball area (MRR, MNP, KR), as well as two populations at the southern extent of F. australis (WG, HB), could represent a geographically separated entities that are genetically distinct from the typical forms found further inland. Genetic differentiation (F_ST) between the Mooball populations and the inland populations of F. australis was not as strong as at inter-species levels, yet PCA results highlighted similar genomic divergence relationships that were also reflected in the phylogenetic topology. These findings provided strong support to delineate the Mooball cluster and two populations in the Wanganui gorge (WG, HB) from the remaining populations of F. australis. This evidence could indicate they be recognised as a subspecies of F. australis. However, this raises a question about the genetic relationships among the other related groups. On the basis of genetic evidence, the classification of one entity as a distinct group with a subspecies (F. australis), while designating F. oraria and the Fontainea entity from Coffs Harbour as separate species might be questionable. Whereas some genetic results support the presence of three distinct lineages within F. australis, F. oraria, and the Coffs Harbour taxon, additional data such as IBD and pairwise F_ST values (genetic differentiation), could potentially challenge the traditional classification scheme and prompt a re-evaluation of how these species are categorised from a genetic perspective.

Genetic structure generally reflected the geographically fragmented and isolated range of F. australis, F. oraria and the Coffs Harbour Fontainea, with signals to indicate limited geneflow, which might be driving genetic drift along the latitudinal distribution of the three groups. This geographic structuring agrees with Brunton et al. (2024), who showed that reciprocal admixture was more frequent in the collected localities of F. australis closest to F. oraria, which included the Mooball cluster as well as WG and HB, to suggest that geneflow among the taxa may have occurred relatively recently. This scenario would likely have involved intermediate, relict populations in the ‘Big Scrub’, an area of subtropical rainforest (Complex Notophyll Vine Forest) that once covered up to 750 km² of the deep, volcanic soils of the Alstonville plateau in northern NSW, from the high inland elevations, to within a few kilometres of the current coastline (Parkes et al. 2012). The fact that expected heterozygosity (H_E) was higher in the smallest population (F. oraria) also supports recent admixed origins. Extensive clearing since the 1840s means that <1% of its original habitat extent remains as small, scattered rainforest fragments (Parkes et al. 2012), and the possibility of contemporary geneflow that includes novel, undiscovered F. australis populations within this region is highly unlikely. However, in the case of Coffs Harbour Fontainea relative to the other taxa, clear genetic differentiation suggests that recent geneflow from its northern congeners is unlikely.

Although some genetic evidence supported the presence of three distinct lineages within F. australis, F. oraria, and the Coffs Harbour taxon, the challenge in classifying each group as a distinct species stems from our broader genetic data, which showed an IBD pattern of genomic divergence. This IBD pattern could suggest that the three taxa represent a species continuum resulting from geographic partitioning, adding complexity to the interpretation of the evolutionary relationships among these Fontainea entities. We observed a north–south genetic cline, tied to the rainforest environment biome envelope where the F. australis populations in the northern extent of its range are genetically clustered. Signals of genetic variation indicated that there is emerging genetic drift driving the divergence of the central populations, which are isolated in refugia of high elevations (NC), deep gorges (HB, WG), and the most easterly situated (Mooball) populations. Fontainea oraria remains isolated within a small, coastal rainforest patch within 1 km of the ocean, and there is a large break of almost 200 km, which includes numerous pronounced geographic barriers, separating the newly discovered Fontainea at Coffs Harbour. Although this result is unsurprising, recent landscape genetic analysis have suggested that environmental and geographic factors play a distinct role in the intraspecific diversity of F. australis and F. oraria (Brunton et al. 2024). Therefore, we could infer that this effect is also a key factor in the genetic divergence of the putative new species F. sp. Coffs Harbour and this entity is a highly diverged geographic race within this complex.

Morphological distinctions for taxonomic delineation

Species identification keys for Fontainea can be difficult to use for field collections because they are mainly based on reproductive features (Jessup and Guymer 1985; Forster 1997; Forster and Van Welzen 1999), which are not always present. To overcome this, leaf traits can offer an objective and independent approach to test the circumscription of taxonomic entities, keeping in mind variation owing to phenotypic expression. Our morphometric results largely corresponded to the current taxonomic treatments in identifying three taxa (albeit preliminary in the case of F. sp. Coffs Harbour). We tested a range of leaf characters that are generally not used in classical systematic treatments; however, metrics such as SLA, PL and DLM have been useful in demonstrating inter-species variation in a number of global (Duminil et al. 2012; Buzatti et al. 2019) and Australian (Wright et al. 2002, 2004) contexts. In this study, morphometric classifications showed that several leaf characters (SLA, DLM and PL) could be used to distinguish among some Fontainea species. Yet, phenological patterns were not well-resolved for all species and may not be regular to highlight the challenge of using only a morphometric approach in a set of plastic traits. This difficulty in distinguishing clear boundaries with leaf traits, therefore, further complicates the task of achieving conclusive species delineation solely on the basis of morphological characteristics used for this study.

For example, our results showed some mixed classifications between F. australis and F. oraria on the basis of the complete set of leaf characters. Leaf traits associated with light capture, such as SLA, were lower in Coffs Harbour Fontainea, which is likely to reflect the variation in local environment among the collection sites for each species. Fontainea australis and F. oraria are located in sites with climatic conditions favourable to plant productivity, such as high annual precipitation and high summer temperatures (Brunton et al. 2023), and showed generally higher values for high SLA and DLM, which signals growth traits in response to these conditions. However, we also found some cases within the classifications of F. australis related to lower SLA, and a handful of examples where all three species were classified together. This complex pattern of morphometric variance was also highlighted by the range of classification analyses that assigned some cases from the Mooball populations with F. oraria, and at various levels (taxon, population and individual) classified F. oraria with F. australis. This may lend support to a subspecies classification of the Mooball populations on the basis of a definition by Stebbins (1950) that suggests only one or several minor differences in characters (in this case SLA, PL, DLM), and an intermediate population (F. oraria) should occur that exhibits a link in character states. In addition, the distinct clustering of these regional populations as a genetic subgroup endorses the recognition of the Mooball population as a geographical race of F. australis that is in line with another subspecies definition by Stace (1991). Where we observed this mixed pattern of species classification may be best explained by the plasticity in leaf shape among the three species. We found numerous leaf forms both within and among the three species from our samples that are reflected even in the small number of species representatives presented in Fig. 2. This clearly showed that each taxon contains a range of leaf shapes from elliptic to oblanceolate, or ovate, which may be difficult to distinguish in the field. The challenge arises when character states overlap to an extent where they cannot be effectively utilised in a key, making it notably difficult to describe and recognise the taxa in the field.

Plasticity in leaf characteristics is distinct in the southern clade of Fontainea, unlike for other species in the genus, aligning with patterns observed in various tropical rainforest species (Wright et al. 2004; Rozendaal et al. 2006). This adaptability, influenced by factors such as climate, light conditions and extreme seasonal variations, is notable in the context of the distribution of this Fontainea clade across an environmentally complex landscape. Despite our data supporting taxonomic groupings based on morphological features within the Fontainea complex, uncertainties persist, particularly regarding certain geographic groups within F. australis–F. oraria. Plastic leaf traits are evident in all three groups, emphasising the need for comparison in a controlled environment (such as a common garden experiment) or when lineages coexist, which is not the case here. So, overall, the predominant signal suggests a pattern of geographic separation and drift in leaf morphology, necessitating further exploration of floral and fruiting characteristics for a conclusive taxonomic determination to support a theory of three distinct taxa.

Overall, our study has highlighted the inherent challenge in defining species boundaries among genetically and morphologically similar taxa, especially in the absence of reproductive material. Even though we found correlations between certain leaf characters and genetically diverged groups, particularly for the Coffs Harbour entity, the evidence remains inconclusive for the separation of F. australis and F. oraria. We suggest here that the pattern of genomic divergence and morphological variability among the three Fontainea taxa represents a continuum structured by prominent breaks in gene flow related to major geographic barriers as well as a complex series of contraction–expansion cycles related to paleoclimatic oscillations (Brunton et al. 2023). Yet, our results supported neither the combination of F. australis with F. oraria in a unified taxon, nor the unique classification of F. australis populations from Mooball and the Wanganui gorge. Nevertheless, we found that there are genetic signals of divergence among both this distinct group of F. australis and F. oraria, which are also represented in the morphometric classifications. Therefore, the complexities of our results do not necessarily support the recognition of a merged species that incorporates the eastern populations of F. australis and F. oraria. However, there is a growing evidence to support a renewed taxonomic revision for these species. Clearer distinction for this sister group would also improve the recovery of evolutionary distinct units that would no doubt offer a range of renewed conservation priorities to enhance the diversity within this narrow, rainforest complex with a significant representation in areas of climate refugia (Brunton et al. 2023).

Conservation of critically endangered and threatened species

There are several Australian examples that have explored the patterns of population divergence using an integrated-data approach to recover the relationships among closely related, threatened species (Robins et al. 2021; Wilson et al. 2022; O’Donnell et al. 2023). These studies highlighted various outcomes, such as subspecies recognition (Robins et al. 2021), designation of new species (O’Donnell et al. 2023) and the merging of some species (Wilson et al. 2022), that all contributed to updated conservation priorities. However, overestimating morphological diversity within species of taxonomic uncertainty can divert attention and has often restricted resources away from populations that are in greater need of conservation focus (Wilson et al. 2022). For the three Fontainea groups in this study, we suggest that they deserve ongoing conservation management to ensure that the species diversity within this already genetically restricted, narrow-range genus is retained. Fontainea oraria is currently undergoing a translocation program aimed at achieving genetic rescue, which has yielded great success in reproductive output and genetic improvements (Rossetto et al. 2023). Although F. australis has a broader distribution and pool of species diversity, the ambiguous relationships between some of their populations and F. oraria stress the importance of maintaining the geographically distinct clusters, prioritising formal protection of more populations through private conservation covenants, or strategically purchasing properties that could expand the current suitable habitat. Habitat protection remains a key conservation focus for all three Fontainea groups in this complex and the lack of formal protection presents the greatest threat to their immediate preservation (Brunton et al. 2023).

Should the new Fontainea entity from Coffs Harbour receive formal description as a new species, it will join five other threatened Australian Fontainea species. This formal description will be crucial in the conservation of this entity, nevertheless its small population size and southern-range extension of the genus support conservation attention. As currently circumscribed, this new population has been assigned a species name, F. sp. Coffs Harbour (National Herbarium of New South Wales 2021), in addition to a provisional Critically Endangered status (PlantNET, see https://plantnet.rbgsyd.nsw.gov.au). This potentially new species is restricted to just two small sites, with a limited number (<5) of mature, reproductive individuals that are subject to ongoing disturbance from road-construction activities. One site contains only a single, mature fruit-producing tree, which is designated for translocation to allow its current habitat to be cleared for a major road upgrade. Importantly, our results indicated that Coffs Harbour Fontainea had low levels of genomic variation compared with F. australis and F. oraria; yet, despite our expectations of a high degree of inter-relatedness, no evidence was found to indicate inbreeding among the limited number of individuals. This suggests that even though the small number of reproductively mature plants represent an adequate number of unrelated individuals or have self-mating limitations, a genetic-rescue program for Fontainea from Coffs Harbour with individuals of low levels of mean kinship will be crucial for its conservation.

Finally, the genetic and morphometric evidence we present has provided information to warrant further taxonomic attention to F. australis and F. oraria. This would determine whether systematic treatments support the signals of genomic and morphometric divergence found in this study or support a single species theory, with distinctions resulting from isolation by distance as result of genetic drift. If a future taxonomic revision determines that F. oraria merely represents a divergent F. australis population, it will not significantly extend the range or genetic diversity of that species. Therefore, the natural population of F. oraria remains a key conservation focus. A number of additional sites from translocated material has been established for this species (Brown et al. 2017); however, more attention is warranted in identifying potential additional translocation areas (Brunton et al. 2023). Similar to F. oraria, material has been captured to develop an ex situ population that can be used to enhance the genetic profile for the Coffs Harbour Fontainea, which is in an extreme risk of extinction, and where appropriate, provide material for translocation to suitable habitats.